With direct digitization of detector pulses, software algorithms are used to calculate the energy of incident photons. Fast high resolution sampling adc's and digital signal processors replace shaping amplifiers and spectroscopy adc's. The difference between the classical analog approach for pulse processing and direct digitization can be summerized as follows. In analog systems, a preamplifier output is filtered by analog means and then digitized to be acquired in a multichannel analyzer, whereas the full digital system will digitize first and then filter by application of algorithms. According to the sampling theorem, both approaches are equivalent. However, using digital signal processing for filtering allows the use of filter functions that cannot be practically realized with analog means. Those digital filter functions promise resolution and throughput close to the theoretical limit. The first commercially available ADSP (Analog to Digital Signal Processor) uses a moving window deconvolution to deconvolve the preamplifier characteristic. All algorithms are calculated in real time, thus there is no dead time added through the computation of an event. A filter function with a trapezoidal impulse response calculates the energy in real time. An adaptive digital trigger allows excellent low energy detection. Another benefit for X-ray applications are very long shaping constants, resulting in 125 eV resolution at 5.9 keV with SiLi detectors. The ADSP consists of an analog linear amplification stage, a 20 MHz sampling adc, circuits for digital signal preprocessing, and four floating point DSPs, performing 240 MFLOP/s. It has been built into a 2-wide NIM module to replace virtually any adc/shaping amplifier combination.